JP2008257327A - Collating device - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide a collating technique that implements accurate collating involving fewer determination errors by a simple structure. <P>SOLUTION: A collating device has a primary collating means for collating a subject's biometrics information against biometrics information stored in a storage means to select a registrant similar to the subject, a secondary collating means for implementing collating after interchanging the biometrics information about the registrant selected by the primary collating means and the subject, and a determination means for determining a registrant collating the subject according to the collating results of the primary collating and secondary collating. The determination means preferably determines that the primary collating result is correct if the subject is selected as the secondary collating result. <P>COPYRIGHT: (C)2009,JPO&INPIT

Description

  The present invention relates to a verification technique using biometric information.

  In recent years, collation / authentication techniques using biometric information such as facial images and fingerprints have become widespread. In such a biometric matching technique, improvement of matching accuracy is a research subject.

  In general, if security is to be ensured, the possibility that the person (registrant) is mistakenly rejected increases, and convenience is sacrificed. In face authentication, the accuracy of matching is affected by environmental changes such as lighting conditions. In view of this, there is known a face authentication device that can adjust parameters related to verification so that the balance between performance and convenience (comfort) of personal authentication and the strength against environmental fluctuation can be changed according to the situation (patent) Reference 1). With this technology, a plurality of modes such as convenience and safety can be switched, and the security strength can be adjusted.

In addition, two collation methods, face collation and voice collation, are prepared. If the matching degree by face matching is sufficiently high, it is regarded as a collation success, but if the matching degree by face matching is not so high, voice collation is performed. Also known is a technique (multimodal biometric authentication) for performing identity verification (Patent Document 2). Thus, by using two types of collation methods, it is possible to improve security and collation accuracy.
JP 2005-149527 A Japanese Patent Laid-Open No. 2004-126913

  However, in the case of the prior art as described above, the following problems have occurred. That is, in order to make it possible to adjust the parameters related to collation, it is necessary to customize the possible parameter values in advance. It is difficult to find the optimum adjustment by this customization work, and it is difficult to improve the matching accuracy by such adjustment. In addition, when a plurality of collation methods are used, there are a problem that the apparatus is increased in size and a cost is increased.

  The present invention has been made in view of the above circumstances, and an object of the present invention is to provide a technique capable of performing collation with higher accuracy by a simple method.

  In order to achieve the above object, a collation apparatus according to the present invention includes a primary collation for collating input data with registered data, and a secondary collation for exchanging the input data and registered data selected as a result of the primary collation. And the registrant that matches the input data is determined based on these two collation results.

  The collation apparatus according to the present invention includes storage means, input means, primary collation means, secondary collation means, and determination means.

The storage means stores biometric information of a plurality of registrants as registration data. The biometric information includes information related to a face, fingerprint, voice, vein, retina, iris, and the like (for example, when using face information, the face image itself or information acquired from the face image such as a face feature amount). included. The input means receives input of biometric information of the person to be verified. When the biological information is a face or a fingerprint, the input means can be configured as an imaging device such as a digital camera. Further, when the biological information is voice, it can be configured as a recording device such as a microphone. In addition, the input means can store the biometric information already acquired by the imaging device or the recording device via a network or various storage media (hard disk device, CD-ROM, DVD-ROM, floppy (registered trademark) disk, It can also be configured to input via various flash memories or the like.

  The primary collation unit collates the biometric information of the input person to be collated with the biometric information of a plurality of registrants stored in the storage unit, and selects a registrant similar to the person to be collated. The primary collating means may be configured to obtain a feature amount from the input biometric information of the person to be collated and compare the feature amount with the biometric information of the registrant to calculate the similarity.

  Then, the secondary verification unit performs verification by exchanging the biometric information of the registrant selected by the primary verification unit and the verification target person. That is, the secondary collation means treats the biometric information of the registrant who has not been selected by the primary collation means and the biometric information of the input collation target as registration data among the registration data stored in the storage means. Then, the secondary verification unit performs verification by treating the biometric information of the registrant selected by the primary verification unit as input data.

  The determination unit determines a registrant that matches the verification target person based on the verification results of the primary verification unit and the secondary verification unit.

  According to such a configuration, the correctness of the primary collation can be confirmed by performing the secondary collation in which the input data and the registered data are exchanged. Accordingly, when the result of the primary collation is incorrect, an erroneous determination can be detected by the secondary collation, so that the collation accuracy can be improved.

  Here, the primary verification means can select one registrant who is most similar to the verification target person as the verification result. In this case, the secondary collation means uses the biometric information stored in the storage means as a registration data obtained by replacing the biometric information of the selected registrant and the biometric information of the person to be collated. Is used as input data. The determining means determines that the collation result by the primary collating means is correct when it is determined that the registrant (input data) selected by the primary collating means matches the person to be collated as the collation result of the secondary collating means. Can be determined. That is, the determination means can determine that the registrant selected by the primary verification means is a registrant that matches the verification target person.

  The collation by the secondary collation means is performed by 1-to-N collation for selecting which registrant among a plurality of registrants stored in the registration data the person having the input biometric information matches. Also good. Further, the collation by the secondary collating means determines whether or not the input biometric information matches the correct answer data by using the biometric information of the person to be collated as correct data and the biometric information of other registrants as incorrect answer data. One-to-one matching may be performed.

  Further, the primary verification unit can select a plurality of registrants having a degree of similarity equal to or higher than a predetermined threshold as a verification result. In this case, the secondary verification unit performs verification for each of the plurality of registrants selected by the primary verification unit by replacing the biometric information of the verification target person. Then, the determination unit can determine which registrant among the plurality of selected registrants matches the person to be collated based on the result of the secondary collation. That is, the determination unit can determine a registrant who is determined to have the highest similarity to the verification target person in the verification by the secondary verification unit as a registrant that matches the verification target person.

In addition, when the biometric information registered in a memory | storage means is one about one registrant, the secondary collation performed by switching a registrant and a collation target person should just be performed once about one registrant. On the other hand, when there are a plurality of pieces of biometric information registered in the storage unit for each registrant, the verification by the secondary verification unit and the determination by the determination unit may be performed as follows.

  First, the secondary collation means performs the secondary collation by replacing only the biometric information having the highest similarity with the biometric information of the person to be collated among the plurality of biometric information of the registrant selected by the primary collation means. be able to. In this case, the secondary verification is performed only once, and the determination by the determination unit is the same as when the registration data is one per registrant.

  Moreover, the secondary verification means can also perform secondary verification for each of the plurality of biometric information of the registrant selected by the primary verification means with the biometric information of the verification target person. In this case, the secondary verification means performs secondary verification for the number of biometric information of the selected registrant. Then, the determination means matches the person to be verified based on the matching result with the highest similarity (which can be said to be a matching result with high reliability) among the results of the secondary matching performed by the secondary matching means. The registrant can be determined. For example, in the secondary matching result with the highest similarity, when it is determined that the registrant selected by the primary matching means and the person to be verified are similar with a similarity equal to or higher than a predetermined threshold, the determining means It can be determined that the registered person is a registered person that matches the person to be verified.

  Further, when the secondary verification unit performs secondary verification for each of the plurality of biometric information of the registrant selected by the primary verification unit with the biometric information of the verification target person, the determination unit is as follows. Thus, a registrant that matches the person to be verified can be determined. In other words, the determination unit can determine a registrant that matches the verification target person based on the average of the results of the multiple secondary verifications by the secondary verification unit. The determination means, for example, takes the average of the similarity between the registrant selected by the primary verification means and the verification target person in the secondary verification result, and if the average value of this similarity is equal to or greater than a predetermined threshold, The registrant selected by the primary verification means can be determined as the registrant that matches the verification target person.

  In the present invention, the primary verification unit and the secondary verification unit perform verification using the same biological information, but it is not necessary to employ the same verification algorithm. For example, the primary verification unit may perform verification focusing on the entire biometric information, and the secondary verification unit may perform verification focusing on only part of the biometric information. Further, the reverse configuration may be adopted.

  For example, when the biometric information is a face image and face collation is performed, the primary collation means performs collation focusing on the characteristics of the entire face, and the secondary collation means only applies to some organs of the face such as the eyes, mouth, and ears. It is possible to perform collation by paying attention to the above.

  Further, a plurality of types of biometric information (such as face and voice) may be stored in the recording unit, and the biometric information focused on by the primary verification unit and the secondary verification unit may be different. For example, the primary collating unit can perform collation focusing on the face, and the secondary collating unit can perform collation focusing on the voice.

  Note that the present invention can be understood as a collation device having at least a part of the above means. The present invention can also be understood as a collation method including at least a part of the above processing, or a program for realizing the method. Each of the above means and processes can be combined with each other as much as possible to constitute the present invention.

For example, in the collation method according to an aspect of the present invention, the information processing apparatus stores a biometric information of a plurality of registrants, an input step of accepting input of biometric information of a collation target person, and the collation target person The biometric information of the registrant and the biometric information of the registrant, a primary verification step of selecting a registrant similar to the verification target person, and the biometric information of the registrant and the verification target person selected in the primary verification step. A secondary collation step for performing collation by exchanging, and a determination step for determining a registrant that matches the collation target person based on the collation results of the primary collation step and the secondary collation step are performed. And

  For example, the collation program which concerns on 1 aspect of this invention WHEREIN: The memory | storage step which stores biometric information of a several registrant with respect to information processing apparatus, the input step which receives the input of biometric information of a collation subject person, A primary collation step of collating the biometric information of the person to be collated with the biometric information of the registrant and selecting a registrant similar to the person to be collated, the registrant selected in the primary collation step, and the person to be collated A secondary collation step for performing collation by exchanging the biometric information, and a determination step for determining a registrant that matches the collation target person based on the collation results of the primary collation step and the secondary collation step. It is characterized by making it.

  According to the present invention, it is possible to perform collation with higher accuracy by a simple method.

  Exemplary embodiments of the present invention will be described in detail below with reference to the drawings.

(First embodiment)
The first embodiment is a face collation device that accepts input of a face image and determines which of the registered users is registered in advance.

<Configuration>
The face matching device 1 according to the present embodiment is configured to include a CPU (Central Processing Unit), a main storage device (RAM), an auxiliary storage device, and the like connected via a bus in terms of hardware. . In this case, the face matching device 1 is realized by executing the program on the CPU.

  FIG. 1 is a diagram illustrating an example of functional blocks of the face matching device 1. The face collating device 1 is loaded with various programs (OS, applications, etc.) stored in the auxiliary storage device and executed by the CPU, whereby the image input unit 2, the face detection unit 3, and the registered image DB 4 are loaded. The primary verification unit 5, the secondary verification unit 6 (including the image management unit 7, the learning unit 8, and the verification unit 9) and the determination unit 10. Further, all or part of the face collation apparatus 1 may be configured as a dedicated chip.

  Next, each functional unit included in the face matching device 1 will be described.

  The image input unit 2 functions as an interface for inputting face image data of the person to be collated into the face collation apparatus 1. The image input unit 2 inputs face image data to the face matching device 1. The image input unit 2 may be configured using any existing technique for inputting face image data to the face matching device 1.

  For example, face image data may be input to the face collation apparatus 1 via a network (such as a LAN or the Internet). Further, a face image may be input to the face collation apparatus 1 from a digital camera, a scanner, a recording device (for example, a hard disk drive), or the like. The face matching device 1 is included in various devices (such as a mobile phone and a PDA (Personal Digital Assistant)) including an imaging device such as a digital camera or an imaging device such as a digital camera. 1 may be input.

The face detection unit 3 detects a human face from the face image input via the image input unit 2. The face detection unit 3 may be configured to detect a face by template matching using a reference template corresponding to the outline of the entire face, for example. The face detection unit 3 may be configured to detect a face by template matching based on face components (eg, eyes, nose, ears). The face detection unit 3 may be configured to detect a region close to the skin color and detect the region as a face. Further, the face detection unit 3 may be configured to perform learning by a teacher signal using a neural network and detect a face-like region as a face. The face detection unit 3 may be realized by applying any other existing technique.

  The registered image DB 4 is configured by an auxiliary storage device such as a hard disk device. The registered image DB 4 stores a registrant's face image. In the registered image DB 4, one registered image can be stored for each registrant, or a plurality of registered images can be stored. In the present embodiment, it is assumed that only one registered image is stored in the registered image DB 4 for each registrant.

  The primary collation unit 5 collates the input face image of the person to be collated with the registered image in the registered image DB 4 to determine which registrant is the person to be collated. The primary collation unit 5 obtains a feature amount from the face image, compares it with the registered image, and calculates a collation score (similarity) between the input image and each registered image. And the primary collation part 5 selects the registrant (henceforth applicable person) who corresponds with a collation subject person based on this collation score (1 to N collation). In the present embodiment, the registrant with the highest matching score is selected as the corresponding person. However, when the highest collation score is smaller than the predetermined threshold, the primary collation unit 5 determines that there is no corresponding person, and determines that the person to be collated is an unregistered person.

  The secondary verification unit 6 performs verification by a method different from that of the primary verification unit 5 in order to improve the verification accuracy. Then, the determination unit 10 determines which registrant matches the verification target person based on the verification results of the primary verification unit 5 and the secondary verification unit 6.

  The secondary collation unit 6 includes an image management unit 7, a learning unit 8, and a collation unit 9. The image management unit 7 manages the input image of the person to be collated input via the image input unit 2 and the registered image stored in the registered image DB 4. The image management unit 7 acquires a registered image and an input image stored in the registered image DB 4. Then, the image management unit 7 replaces the registered image of the registrant (corresponding person) selected by the primary verification unit 5 among the registered images with the input image, and passes these face images to the learning unit 8. In addition, the image management unit 7 passes the registered image of the corresponding person to the collation unit 9 as input data.

  As described above, the learning unit 8 receives from the image management unit 7 a set of face images obtained by replacing the registered image of the corresponding person among the registered images stored in the registered image DB 4 with the input image. The learning unit 8 performs learning processing based on these face images. The learning method may be performed by any existing technique such as PCA, template, or SVM.

  The collation unit 9 performs face collation using the face image of the corresponding person passed from the image management unit 7 as input data and the learning result of the learning unit 8 as a learning data set. The matching unit 9 determines which person in the learning data set matches the corresponding person as input data. Specifically, the collation unit 9 compares the facial feature amounts to calculate a collation score, similarly to the primary collation unit 5, and determines a person who matches the corresponding person based on the collation score.

The determination unit 10 determines which registrant matches the verification target person input via the image input unit 2 based on the verification results of the primary verification unit 5 and the secondary verification unit 6. If it is determined as a matching result by the secondary matching unit 6 that the corresponding person matches the matching target person, the determining unit 10 determines that the matching result by the primary matching unit 5 is correct. That is, in this case, the determination unit 10 determines that the person to be verified matches the corresponding person. Further, when a result that a person other than the person to be collated matches the corresponding person is obtained as a collation result by the secondary collation unit 6, the determination unit 10 determines that the collation by the primary collation unit 5 is not correct. That is, the determination unit 10 determines that there is no registrant that matches the verification target person, and the verification target person is an unregistered person.

<Operation example>
2 and 3 are flowcharts showing an example of the operation of the face collation apparatus 1. FIG. 4 is a diagram for explaining a collation method of the face collation apparatus 1.

  First, with reference to FIG. 2, the primary collation process by the primary collation part 5 is demonstrated. FIG. 2 is a flowchart mainly showing the flow of collation processing by the primary collation unit 5. First, the face collation apparatus 1 accepts input of a face image (collation image) of the person to be collated via the image input unit 2 (S10). Then, the face detection unit 3 detects a face from the collation image (S11). Next, the primary collation unit 5 collates the face detected from the collation image with the registered image in the registered image DB 4 and selects a registrant having the highest similarity with the collation target person (S12). It is determined whether or not the similarity is higher than a predetermined threshold (S13). If the similarity is higher than the threshold (S13-YES), it is determined that the registrant is a registrant (corresponding person) that matches the person to be verified. (S14). Whether or not this collation result is correct is confirmed by secondary collation. On the other hand, when the similarity is equal to or less than a predetermined threshold (S13-NO), it is determined that the person to be verified is an unregistered person. In this case, it is concluded that the person is an unregistered person without confirmation by the secondary verification unit 6.

  Here, as shown in FIG. 4, it is assumed that four persons A to D are registered as registrants, and it is determined that the collation image is Mr. B as a result of the primary collation.

  Next, with reference to FIGS. 3 and 4, the secondary verification process by the secondary verification unit 6 will be described. FIG. 3 is a flowchart mainly showing the flow of collation processing by the secondary collation unit 6. First, the image management unit 7 replaces the collation image and the registered image of the person (Mr. B) (S20). Then, the learning unit 8 performs a learning process based on the face images of a total of four people including the person to be collated and Mr. A, Mr. C, and Mr. D (S21). Also, the image management unit 7 inputs the registered image of the person (Mr. B) as input data to the verification unit 9 (S22). The collation unit 9 performs 1-to-N collation using the learning result described above, and determines which person Mr. B matches (S23). Thereafter, the determination unit 10 determines whether or not the face image (collation image) of the person to be collated has been selected as the person who matches Mr. B (24). When the face image of the person to be collated is selected (S24-YES), the determination unit 10 determines that the collation by the primary collation unit 5 is correct. Conversely, when the face image of the person to be collated is not selected (S24-NO), it is determined that the collation by the primary collation unit 5 is not correct, and the person to be collated is determined to be an unregistered person. The case where the face image of the person to be collated is not selected includes the case where another registrant is selected and the case where there is no corresponding person.

<Operation / Effect of Embodiment>
In the face collation apparatus 1 according to the present embodiment, the result of the primary collation is confirmed by the secondary collation performed by exchanging the collation image and the image of the registrant as the result of the primary collation. Therefore, when the result of the primary collation is incorrect, the error can be detected by the secondary collation, so that erroneous discrimination can be reduced. Therefore, the face matching device 1 improves the matching accuracy.

  Further, in the face collation apparatus 1 according to the present embodiment, the same collation called face collation processing is performed in either the primary collation or the secondary collation. Therefore, the configuration can be simplified as compared with a multi-modal collation apparatus that uses a plurality of collation methods in combination, and an increase in size and cost of the apparatus can be avoided.

(Second Embodiment)
In the first embodiment, the one-to-N collation is performed as the collation processing by the secondary collation unit 6. In the present embodiment, the second collation is performed by one-to-one collation.

  FIG. 5 is a flowchart showing the flow of the secondary collation process in the present embodiment, and FIG. 6 is a diagram for explaining the secondary collation process.

  For the primary verification, as in the first embodiment, a registrant (corresponding person) that matches the input verification target person is selected. In the secondary collation process, the image management unit 7 replaces the collation image with the registered image of the corresponding person (S30). Then, the learning unit 8 learns using the verification images as correct answer data and the registered images other than the corresponding person (in the example of FIG. 6, three images of Mr. A, Mr. C, and Mr. D) as incorrect answer data. Further, the image management unit 7 inputs the registered image of the person (Mr. B) as input data to the verification unit 9 (S32). Using the learning result, the matching unit 9 performs one-to-one matching as to whether or not the registered image of the corresponding person as input data belongs to correct data (S33).

  The determination unit 10 checks whether or not the corresponding person (Mr. B) is determined to belong to the correct answer data as a verification result by the secondary verification unit 6 (S34). When it is determined that the corresponding person belongs to the correct data class (34-YES), it is determined that the collation by the primary collation unit 5 is correct. Conversely, if it is determined that the corresponding person belongs to the class of incorrect data (S34-NO), it is determined that the collation by the primary collation unit 5 is not correct, and the collation target person is determined to be an unregistered person. .

  Also in the face collation apparatus 1 according to the present embodiment, the same effects as in the first embodiment can be obtained.

(Third embodiment)
In the above embodiment, in both the primary collation process and the secondary collation process, the feature amount of the entire face is acquired and the collation score is calculated. On the other hand, in this embodiment, the algorithm for calculating the matching score is different between the primary matching process and the secondary matching process. That is, the face part of interest is different in the primary matching process and the secondary matching process.

  For example, as shown in FIG. 7, in the primary collation, the collation score is calculated by focusing on the entire face, whereas in the secondary collation, the collation score is calculated by focusing only on the eyes. Also good. Also, for example, as shown in FIG. 8, in the primary collation, the collation score is calculated by focusing on the entire face, whereas in the secondary collation, the collation score is calculated by focusing only on the mouth. You may do it. As described above, in the secondary matching, a matching algorithm that focuses only on a part of the face (an organ such as the eyes, nose, mouth, ear, or a combination thereof or a positional relationship thereof) can be employed. Further, some organs may be focused on in the primary matching, and the entire face may be focused on in the secondary matching. In addition, it is also possible to perform collation focusing on a part of the face in both the primary collation and the secondary collation, such as focusing on eyes only in the primary collation and focusing on the mouth in the secondary collation.

  When focusing on only a part of the organs of the face, for example, when wearing a mask, the accuracy of matching can be improved by not using the hidden mouth as a target for eye comparison. Moreover, when wearing sunglasses, the accuracy of verification can be improved by using the mouth as a comparison target and not using the hidden eyes for verification. Thus, collation accuracy is improved by performing two types of collation for the face image in which a part of the face is hidden and targeting the entire face and some organs (not hidden). Can do.

  In addition, wearing of a mask or sunglasses can be determined by a technique such as template matching. Further, it is possible to detect that some organs are hidden in the face image by examining the positional relationship of the feature points with respect to the difference in face detection.

  In the present embodiment, information related to a face is used as biometric information, and the organ of the face to which attention is paid is changed between primary verification and secondary verification, but the same configuration can be adopted when other biometric information is used. it can. Further, in the primary verification and the secondary verification, the verification may be performed by paying attention to different biological information.

(Fourth embodiment)
In the description of the above embodiment, the case where only one registered image is stored in the registered image DB 4 for each registrant has been described. In the present embodiment, a case will be described in which a plurality of registered images (n) are stored for each registrant in the registered image DB 4.

  The primary verification process performed by the primary verification unit 5 is the same as that in the above embodiment, and the registrant having the highest similarity with the verification image is selected as the corresponding person.

  Hereinafter, the secondary verification process performed by the secondary verification unit 6 in the present embodiment will be described with reference to FIG. FIG. 9 is a flowchart showing the flow of the secondary collation process.

  First, the secondary verification unit 6 substitutes 1 for the variable i (S40). Next, the image management unit 7 replaces the collation image with the i-th registered image of the corresponding person (S41). Then, the learning unit 8 performs a learning process using, as learning data, a set of face images obtained by replacing the i-th registered image of the corresponding person with a matching image among the registered images in the registered image DB 4 (S42). Next, the image management unit 7 passes the i-th registered image of the corresponding person to the verification unit 9 as input data (S43). The collation unit 9 collates the i-th registered image of the corresponding person as input data using the learning result (S44). Then, the collation unit 9 calculates the similarity Si between the i-th registered image of the corresponding person and the collation image (S45).

  Next, it is determined whether or not the processing has been completed for all images of the corresponding person. That is, it is determined whether the variable i is smaller than n (S46). If the variable i is smaller than n, that is, there is an image that has not been processed (S46-YES), the variable i is incremented (S47), and the above processing is repeated for the next registered image.

  When the above process is completed for all n registered images of the corresponding person (S46-NO), it is determined whether the maximum value of the similarity Si (i = 1 to n) is equal to or greater than a predetermined threshold (S48). ). If the maximum value of the similarity score Si is equal to or greater than the threshold value (S48-YES), it is determined that the result of the primary collation is correct. When the maximum value of similarity score Si is smaller than the threshold value (S48-NO), it is determined that the result of the primary verification is an error, and the verification target person is determined to be an unregistered person.

  According to such a configuration, it is possible to cope with a case where a plurality of face images are registered for each registrant in the registered image DB 4.

  In addition, when a plurality of face images are registered for each registrant, a method other than the above can be adopted as the method of secondary verification. Hereinafter, another example of the secondary collation will be described.

  For example, instead of comparing the maximum similarity Si with the threshold value in step S48, the average value of Si and the threshold value can be compared.

  Further, for example, the registered images to be replaced in the secondary collation can be only a predetermined number of registered images determined to have high similarity in the primary collation. For example, only registered images determined to have the highest degree of similarity may be configured to perform learning / collation processing by replacing them with collation images. Further, for example, it is possible to adopt a configuration in which learning / collation processing is performed for a predetermined number of registered images from the one with a higher degree of similarity, replacing the collation images.

(Fifth embodiment)
In the embodiment described above, one registrant who matches the person to be collated is selected by the primary collation process, and whether the primary collation is correct is confirmed by the secondary collation process. On the other hand, the present embodiment is characterized in that a plurality of registrants who are highly likely to match the person to be collated are selected by one collation process and narrowed down by the secondary collation process.

  FIG. 10 is a flowchart showing the flow of the primary verification process by the primary verification unit 5 in the present embodiment. The face collation apparatus 1 accepts input of the face image of the person to be collated via the image input unit 2 (S50). Then, the face detection unit 3 detects a face from the input face image (S51). The primary collation unit 5 collates the face detected from the collation image with the registration image in the registration image DB 4 (S52), and selects the top n registrants with high similarity to the collation target person as candidates ( S53). However, the primary verification unit 5 may set a threshold value for the similarity, and may not include a registrant who has only a similarity less than or equal to the threshold as a candidate. In this case, there may be fewer than n registrants selected as candidates. A registrant who has a similarity within a predetermined value from the maximum similarity can also be set as a candidate.

  Next, the secondary verification process will be described with reference to FIG. FIG. 11 is a flowchart showing the flow of secondary verification processing by the secondary verification unit 6. First, the secondary verification unit 6 substitutes 1 for the variable i (S60). Then, the image management unit 7 replaces the collation image with the registered image of the i-th candidate (S61). Next, the learning unit 8 performs a learning process using the face image after the replacement as learning data (S62). Then, the image management unit 7 passes the registered image of the i-th candidate as input data to the matching unit 9 (S63). The collation unit 9 collates the registered image of the i th candidate, which is input data, using the learning result (S64). Then, the matching unit 9 calculates the similarity Si between the registered image of the i-th candidate and the matching image (S65).

  Next, it is determined whether the above processing has been completed for all candidates. That is, it is determined whether or not the variable i is smaller than n (S66). If the variable i is smaller than n, that is, there is a candidate who has not been processed (S66-YES), the variable i is incremented (S67), and the above processing is repeated for the next candidate.

  When the above processing is completed for all n candidates (S66-NO), the determination unit 10 determines that the candidate having the highest similarity Si is a registrant that matches the collation image (S68). .

  According to such a configuration, registrants that match the verification target person by primary verification are narrowed down to some extent, and registrants that match by secondary verification are determined, so that the verification accuracy can be improved. In particular, if the similarity (matching score) of a plurality of registrants is high as a result of the primary collation, it is possible to reduce misjudgment by performing confirmation by the secondary collation process.

(Other)
In the above description, a configuration is adopted in which a face image is input to the face matching device. However, it is only necessary that biometric information used for matching is input to the face matching device. Therefore, when a facial feature value is used for matching, a configuration in which a facial feature value is input to the face matching device instead of the face image itself may be employed.

  In the above description, the face collation apparatus using the face image as the biometric information has been described. However, the present invention is not limited only to the face collation. In addition to the face image, the fingerprint, voice, vein, retina, Any biological information such as irises or their feature quantities can be used.

  In addition, the biometric information does not have to be the same between the primary verification and the secondary verification. For example, a face image may be used for primary verification, and voice may be used for secondary verification.

It is a figure which shows the example of a functional block of the face collation apparatus which concerns on 1st Embodiment. It is a flowchart which shows the operation example of the primary collation process in 1st Embodiment. It is a flowchart which shows the operation example of the secondary collation process in 1st Embodiment. It is a conceptual diagram explaining the method of face collation in 1st Embodiment. It is a flowchart which shows the operation example of the secondary collation process in 2nd Embodiment. It is a conceptual diagram explaining the method of face collation in 2nd Embodiment. It is a conceptual diagram explaining the method of face collation in 3rd Embodiment. It is a conceptual diagram explaining the method of face collation in 3rd Embodiment. It is a flowchart which shows the operation example of the secondary collation process in 4th Embodiment. It is a flowchart which shows the operation example of the primary collation process in 5th Embodiment. It is a flowchart which shows the operation example of the secondary collation process in 5th Embodiment.

Explanation of symbols

DESCRIPTION OF SYMBOLS 1 Face collation apparatus 2 Image input part 3 Face detection part 4 Face image DB
5 Primary collation unit 6 Secondary collation unit 7 Image management unit 8 Learning unit 9 Collation unit 10 Determination unit

Claims (12)

Storage means for storing biometric information of a plurality of registrants;
Input means for accepting input of biometric information of the person to be verified;
Primary collation means for collating the biometric information of the person to be collated with the biometric information stored in the storage means, and selecting a registrant similar to the person to be collated;
Secondary collation means for performing collation by exchanging biometric information of the registrant selected by the primary collation means and the person to be collated;
A determination unit that determines a registrant that matches the verification target person based on a verification result of the primary verification unit and the secondary verification unit;
The collation apparatus characterized by having. The primary verification means selects a registrant that is most similar to the verification target person,
The determination unit determines a registrant selected by the primary verification unit as a registrant that matches the verification target person when the verification result of the secondary verification unit is the verification target person. The collation apparatus according to claim 1. The primary collation means selects a plurality of registrants having a similarity equal to or higher than a predetermined threshold,
The secondary verification unit performs verification for each of a plurality of registrants selected by the primary verification unit with the biometric information of the verification target person,
The determination means determines a registrant determined as having the highest similarity to the verification target person in the verification by the secondary verification means as a registrant that matches the verification target person. Item 4. The verification device according to item 1.   The collation apparatus according to claim 1, wherein the collation by the secondary collation unit is a one-to-N collation.   The collation apparatus according to claim 1, wherein the collation by the secondary collation unit is a one-to-one collation. The storage means stores a plurality of biometric information per registrant,
The secondary collating means replaces and collates only the biometric information having the highest degree of similarity with the biometric information of the person to be collated among the plurality of biometric information of the registrant selected by the primary collating means. The collation apparatus in any one of Claims 1-5. The storage means stores a plurality of biometric information per registrant,
The secondary collating unit collates each of the plurality of biometric information of the registrant selected by the primary collating unit with the biometric information of the verification target person,
The said determination means determines the registrant who corresponds with the said collation subject person based on the collation result with the highest similarity by the said secondary collation means. The Claim 1 characterized by the above-mentioned. Verification device. The storage means stores a plurality of biometric information per registrant,
The secondary collating unit collates each of the plurality of biometric information of the registrant selected by the primary collating unit with the biometric information of the verification target person,
The collation according to claim 1, wherein the determination unit determines a registrant that matches the collation target person based on an average of a plurality of collation results by the secondary collation unit. apparatus.   The collation apparatus according to claim 1, wherein the biological information is a face image. The primary verification means performs verification for the entire face,
The collation apparatus according to claim 9, wherein the secondary collation unit collates only a part of an organ of a face. Information processing device
A storage step for storing biometric information of a plurality of registrants;
An input step for receiving input of biometric information of the person to be verified;
A primary collation step of collating the biometric information of the person to be collated with the biometric information of the registrant and selecting a registrant similar to the person to be collated;
A secondary collation step of performing collation by exchanging the biometric information of the registrant selected in the primary collation step and the collation subject;
A determination step of determining a registrant that matches the verification target person based on the verification result of the primary verification step and the secondary verification step;
The verification method characterized by performing. For information processing equipment
A storage step for storing biometric information of a plurality of registrants;
An input step for receiving input of biometric information of the person to be verified;
A primary collation step of collating the biometric information of the person to be collated with the biometric information of the registrant and selecting a registrant similar to the person to be collated;
A secondary collation step for performing collation by exchanging biometric information of the registrant selected in the primary collation step and the collation target person;
A determination step of determining a registrant that matches the verification target person based on the verification result of the primary verification step and the secondary verification step;
The collation program characterized by performing.

Images